Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Complex Predicates in Urdu
Tafseer Ahmed
Universitaet Konstanz
July 2011
Outline
Complex Predicates
Lights verbs used in Complex Predicates
Complex Predicates and Verb Classes
Conclusion and Future Work
Outline
Complex Predicates
Lights verbs used in Complex Predicates
Complex Predicates and Verb Classes
Conclusion and Future Work
Complex Predicates
I around 700 simple verbs in Urdu.
I many more complex predicates (Butt 1993)
I possible combinations: Adj + V, N + V, PP + V
nAdiyA=nENadya=Erg
yAsin=kOYasin=Acc
yAdmemory
ki-yAdo-Perf.M.Sg
’Nadya remembered Yasin.’ N+V complex predicate
nAdiyA=nENadya=Erg
mEztable.F.Sg
sAfclean
kIdo-Perf.F.Sg
’Nadya cleans a/the table.’ Adj+V complex predicate
Urdu Verbs and Agreement
Before the discussion of different types of potential complex predicates, we need to see the
agreement patterns of simple verbs and objects in Urdu/Hindi.
I When the subject is not marked by a case marker (ergativeetc.), then the verb agrees with the subject.
yAsInYasin.M.Sg
kitAbbook.F.Sg
paRH-tAread-Impf.M.Sg
heAux.Pres
’Yasin reads the book.’
Urdu Verbs and Agreement
I When the subject is marked by a case marker and the objectis unmarked, then the verb agrees with the object.
yAsIn=nEYasin.M.Sg=Erg
kitAbbook.F.Sg
paRH-Iread-Perf.F.Sg
’Yasin read the book.’
I When both the subject and the object are marked by casemarkers, then the verb has default (masculine singular)gender.
nAdiyA=nENadya.F.Sg=Erg
kitAb=kObook.F.Sg=Acc
paRH-Aread-Perf.M.Sg
’Nadya read the book.’
Urdu Verbs and Agreement
I When the subject is marked by a case marker and the objectis unmarked, then the verb agrees with the object.
yAsIn=nEYasin.M.Sg=Erg
kitAbbook.F.Sg
paRH-Iread-Perf.F.Sg
’Yasin read the book.’
I When both the subject and the object are marked by casemarkers, then the verb has default (masculine singular)gender.
nAdiyA=nENadya.F.Sg=Erg
kitAb=kObook.F.Sg=Acc
paRH-Aread-Perf.M.Sg
’Nadya read the book.’
Urdu Verbs and Agreement
I When the subject is marked by a case marker and the objectis unmarked, then the verb agrees with the object.
yAsIn=nEYasin.M.Sg=Erg
kitAbbook.F.Sg
paRH-Iread-Perf.F.Sg
’Yasin read the book.’
I When both the subject and the object are marked by casemarkers, then the verb has default (masculine singular)gender.
nAdiyA=nENadya.F.Sg=Erg
kitAb=kObook.F.Sg=Acc
paRH-Aread-Perf.M.Sg
’Nadya read the book.’
Types of (potential) Complex Predicates
1. The light verb does not agree with the noun.
2. The light verb may agree with the noun.
2.1 The noun does not have modifiers.2.2 The noun may have modifiers.
Types of (potential) Complex Predicates
I Class 1: The light verb does not agree with the noun.I anjum=nE
Anjum=ErgnAdiyA=kONadya=Acc
yAdmemory
ki-yAdo-Perf.M.Sg
’Anjum remembered Nadya.’
I yAd ’memory’ is feminine in Urdu. However, it does not takepart in agreement scheme in yAd+kar complex predicates.
I Class 2: The light verb may agree with the noun.
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
behesdebate.F.Sg
k-Ido-Perf.F.Sg
’Anjum argued with Nadya.’
Types of (potential) Complex Predicates
I Class 1: The light verb does not agree with the noun.I anjum=nE
Anjum=ErgnAdiyA=kONadya=Acc
yAdmemory
ki-yAdo-Perf.M.Sg
’Anjum remembered Nadya.’
I yAd ’memory’ is feminine in Urdu. However, it does not takepart in agreement scheme in yAd+kar complex predicates.
I Class 2: The light verb may agree with the noun.
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
behesdebate.F.Sg
k-Ido-Perf.F.Sg
’Anjum argued with Nadya.’
Types of (potential) Complex Predicates
I Class 2.2: The noun of N+V complex predicates may havemodifiers.
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
savAlquestion.M.Sg
ki-yAdo-Perf.M.Sg
’Anjum asked Nadya.’
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
kAIseveral
accHEgood.M.Pl
savAlquestion.M.Pl
ki-Edo-Perf.M.Pl
’Anjum asked Nadya several good questions.’
Types of (potential) Complex Predicates
I Class 2.2: The noun of N+V complex predicates may havemodifiers.
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
savAlquestion.M.Sg
ki-yAdo-Perf.M.Sg
’Anjum asked Nadya.’
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
kAIseveral
accHEgood.M.Pl
savAlquestion.M.Pl
ki-Edo-Perf.M.Pl
’Anjum asked Nadya several good questions.’
Does class 2 contain real complex predicate?
I Yes. Because the noun in these N+V sequences introduces anargument. (The verb kar ’do’ has two arguments.)
I N+kar ’do’ with dative kO marked argument: hidAyat’instruction/advice’, pESkaS ’offer’, ImEl ’email’
I N+kar ’do’ with sE (instrument) marked argument: SAdI’marriage’, laRAI ’fight’
I N+kar ’do’ with par (locative) marked argument: hamlA’attack’, EtrAz ’objection’
I N+kar ’do’ with k- (genitive) marking the noun: intizAr ’wait’,a ”
Does class 2 contain real complex predicate?
I Yes. Because the noun in these N+V sequences introduces anargument. (The verb kar ’do’ has two arguments.)
I N+kar ’do’ with dative kO marked argument: hidAyat’instruction/advice’, pESkaS ’offer’, ImEl ’email’
I N+kar ’do’ with sE (instrument) marked argument: SAdI’marriage’, laRAI ’fight’
I N+kar ’do’ with par (locative) marked argument: hamlA’attack’, EtrAz ’objection’
I N+kar ’do’ with k- (genitive) marking the noun: intizAr ’wait’,a ”
Does class 2 contain real complex predicate?
I Yes. Because the noun in these N+V sequences introduces anargument. (The verb kar ’do’ has two arguments.)
I N+kar ’do’ with dative kO marked argument: hidAyat’instruction/advice’, pESkaS ’offer’, ImEl ’email’
I N+kar ’do’ with sE (instrument) marked argument: SAdI’marriage’, laRAI ’fight’
I N+kar ’do’ with par (locative) marked argument: hamlA’attack’, EtrAz ’objection’
I N+kar ’do’ with k- (genitive) marking the noun: intizAr ’wait’,a ”
Does class 2 contain real complex predicate?
I Yes. Because the noun in these N+V sequences introduces anargument. (The verb kar ’do’ has two arguments.)
I N+kar ’do’ with dative kO marked argument: hidAyat’instruction/advice’, pESkaS ’offer’, ImEl ’email’
I N+kar ’do’ with sE (instrument) marked argument: SAdI’marriage’, laRAI ’fight’
I N+kar ’do’ with par (locative) marked argument: hamlA’attack’, EtrAz ’objection’
I N+kar ’do’ with k- (genitive) marking the noun: intizAr ’wait’,a ”
More examples of class 2 complex predicates
just for revision. The sentences have an extra (third) argument.
I anjum=nEAnjum=Erg
nAdiyA=parNadya=Inst
EtrAzseveral
ki-yAobjection.M.Sg
do-Perf.M.Sg
’Anjum objected to Nadya.’
I anjum=nEAnjum=Erg
nAdiyA=sENadya=Inst
kAIseveral
accHEgood.M.Pl
savAlquestion.M.Pl
ki-Edo-Perf.M.Pl
’Anjum asked Nadya several good questions.’
Previous Work
I Mohanan (1993,1994) says that N+V sequences in which theverb agrees with the noun have a light verb.
I The noun is the syntactic object as well as part of thecomplex predicate.
I It is possible because these are related to different levels ofrepresentation (ARG STR and GF STR).
Modeling
I Urdu ParGram (PARallel GRAMmer), Universitaet Konstanz,Germany
I Grammar rules written using Lexical Functional Grammar(LFG) framework
I The N+V complex predicates of class 1 were alreadyimplemented in Urdu ParGram.
Modeling
nAdiyA=nENadya
kAmErg
SurUwork
kiyAstart do-Perf.M.Sg
Nadya started the work.F (Functional) Structure of the above sentence is:
Modeling
nAdiyA=nENadya
kAmErg
SurUwork
kiyAstart do-Perf.M.Sg
Nadya started the work.
F (Functional) Structure: A rough sketch
PRED kar <nAdiyA,SurU<kAm>>SUBJ nAdiyAOBJ kAm
SurU ’start’ is not the object.
Modeling
I We decided that class 2 have complex predicates.
I Thanks to Tracy King in the implementation of the grammerrule.
ModelingF-structure of
biccHU=nEscorpion=Erg
meNDak=sEfrog=Inst
behesdebate.F.Sg
k-Ido-Perf.F.Sg
The scorpion argued with the frog.’
Modeling
F-structure of
biccHU=nEscorpion=Erg
meNDak=sEfrog=Inst
behesdebate.F.Sg
k-Ido-Perf.F.Sg
The scorpion argued with the frog.’
F (Functional) Structure: A rough sketch
PRED kar<biccHU,behes<mENDak>>SUBJ biccHUOBJ behesOBL mENDak
behes ’debate’ is the object as well as part of the complexpredicate.
Modeling
biccHU=nEscorpion=Erg
meNDak=sEfrog=Inst
kAIseveral
accHIgood
bAtENdiscussion
k-INdo-Perf
The scorpion discussed several good things with the frog.’
bAT ’thing/matter’ has modifiers.
Outline
Complex Predicates
Lights verbs used in Complex Predicates
Complex Predicates and Verb Classes
Conclusion and Future Work
Light Verbs used in Complex Predicates
I Commonly used light verbs in N+V complex predicates:I kar ’do’I he ’be’I hO ’become’I rakH ’put’I rah ’stay’.
I These light verbs are related with aspect.
Light Verbs used in Complex Predicates
I us=nE3SG=Erg
sabaqlesson
yAdmemory
kiyAdo.Perf.M.Sg
He remembered/learnt the lesson.’
I us=kO3SG=Dat
sabaqlesson
yAdmemory
hU-Abecome-Perf.M.Sg
He remembers the lesson.’
I us=nE3SG=Erg
sabaqlesson
yAdmemory
rakH-Akeep-Perf.M.Sg
He kept the lesson remebered.’
Light Verbs used in Complex Predicates
I us=kO3SG=Dat
sabaqlesson
yAdmemory
hebe.Pres
He remembers the lesson.’
I us=kO3SG=Dat
sabaqlesson
yAdmemory
rah-Astay-Perf
He remembered the lesson.’
Hence, we should not list the N+V complex predicates
as unrelated combinations like N1+V1, N1+V2, N2+V1,
...
We should focus on the noun (or adjective) part of
the complex predicate and find which light verbs
comes with this noun (or adjective).
Light Verbs used in Complex Predicates
I us=kO3SG=Dat
sabaqlesson
yAdmemory
hebe.Pres
He remembers the lesson.’
I us=kO3SG=Dat
sabaqlesson
yAdmemory
rah-Astay-Perf
He remembered the lesson.’
Hence, we should not list the N+V complex predicates
as unrelated combinations like N1+V1, N1+V2, N2+V1,
...
We should focus on the noun (or adjective) part of
the complex predicate and find which light verbs
comes with this noun (or adjective).
Light Verbs used in Complex Predicates
There are other light verbs e.g A ’come’ and dE’give’ that forms N+V complex predicates. However,
these are not productive as the set of kar ’do’/ hO
’become’ verbs. These are discussed briefly at the
end of the presentation.
Outline
Complex Predicates
Lights verbs used in Complex Predicates
Complex Predicates and Verb Classes
Conclusion and Future Work
Verb Classes and Syntax
I Ahmed and Butt (2011) This work is done before I realized that there are
two more "aspectual" light verbs rakH and rah beside the light verbs presented in the
following.
I Commonly used light verbs in N+V complex predicates: kar’do’, he ’be’ and hO ’become’.
I Every noun does not occur with each of these light verbs.
I We follow Levin (1993)’s classic assumption that semanticpredicational classes can be identified on the basis of a studyof the syntactic contexts the predicates occur in.
Verb Classes and Syntax
I Ahmed and Butt (2011) This work is done before I realized that there are
two more "aspectual" light verbs rakH and rah beside the light verbs presented in the
following.
I Commonly used light verbs in N+V complex predicates: kar’do’, he ’be’ and hO ’become’.
I Every noun does not occur with each of these light verbs.
I We follow Levin (1993)’s classic assumption that semanticpredicational classes can be identified on the basis of a studyof the syntactic contexts the predicates occur in.
Methodology
I List of first 45 nouns occuring in N-V combination with eitherof the light verbs kar ’do’, he ’be’ and hO ’become’ in a POStagged corpus compiled by CRULP.
I A full set of combinatorial (im)possibilities of these nounscompiled by using native speaker judgment.
I An analysis of the resulting patterns identified distinctsemantically coherent classes/classification patterns.
I Pertinent semantic factors appear to be stative vs. eventivenouns, agentivity vs. experiencer verbs (psych predications)and the licensing of a dative recipient.
Methodology
I List of first 45 nouns occuring in N-V combination with eitherof the light verbs kar ’do’, he ’be’ and hO ’become’ in a POStagged corpus compiled by CRULP.
I A full set of combinatorial (im)possibilities of these nounscompiled by using native speaker judgment.
I An analysis of the resulting patterns identified distinctsemantically coherent classes/classification patterns.
I Pertinent semantic factors appear to be stative vs. eventivenouns, agentivity vs. experiencer verbs (psych predications)and the licensing of a dative recipient.
Methodology
I List of first 45 nouns occuring in N-V combination with eitherof the light verbs kar ’do’, he ’be’ and hO ’become’ in a POStagged corpus compiled by CRULP.
I A full set of combinatorial (im)possibilities of these nounscompiled by using native speaker judgment.
I An analysis of the resulting patterns identified distinctsemantically coherent classes/classification patterns.
I Pertinent semantic factors appear to be stative vs. eventivenouns, agentivity vs. experiencer verbs (psych predications)and the licensing of a dative recipient.
Verb Classes
I Class APysch verbs: yaqIn ’belief’, piyAr ’love’
Subj (Experiencer) Obj (Theme) N+kar ’do’Subj (Experiencer)=Dat Obj (Theme) N+ hO ’become’Subj (Experiencer)=Dat Obj (Theme) N+ he ’be’
I Class BMain pattern (38/45): ijAd ’invention’, tAmir ’construction’
Subj (Agent) Obj (Theme) N+kar ’do’Subj (Theme) N+hO ’become’/?he ’be’*Subj=Dat Obj N+hO ”become’/he ’be’
Verb Classes
I Class A
nAdiyA=kONadya.F.Sg=Dat
kahanistory.F.Sg.Nom
yAdmemory
hu-I/hebe.Perf-3.F.Sg/be.Pres-3.F.Sg
‘Nadya remembers/knows a/the story.’
I Class B
*(nAdiyah=kO)Nadya.F.Sg=Dat
makAnhouse.M.Sg.Nom
taa2mIrconstruction
hu-A/?hebe.Perf-3.M.Sg/be.Pres-3.M.Sg
‘A/the house got constructed./ A/the house is constructed.’
Other Verb Classes
Class C
I This class allows dative subject like class A. However, it doesnot allow N+hO ’become’ construction.
I nAdiyA=kONadya.F.Sg=Dat
yAsIn=kAYasin.M.Sg=Gen
intizArwaiting
he/*hu-Abe.Pres-3.M.Sg/be.Perf-3.M.Sg
‘Nadya waited for Yasin.’
I hO ’become’ does not work with these nouns because thesubject is too agentive to be felicitous as the undergoer of abecome predication.
Outline
Complex Predicates
Lights verbs used in Complex Predicates
Complex Predicates and Verb Classes
Conclusion and Future Work
Conclusions
I We find Urdu N+V complex predicates in which the noun partof the complex predicate also acts as the syntactic object andthe verb might agree with it.
I The noun of such complex predicates also allows modifiers.
I We find five light verbs kar and hO etc. that frequently follownouns in N+V complex predicate for ”aspectual” reasons.
I We find semantic classes of the nouns formed by(im)permissibility of the above mentioned light verbs.
Future Work
I There are other light verbs e.g A ’come’ and dE ’give’ thatforms N+V complex predicates. However, these are notproductive as the set of kar ’do’/ hO ’become’ verbs. Weguess that these are also related to semantic classes of noun,but it needs further investigation.
I We need to define rules to distinguish the borderline between:(i) main verb and canonical object(ii) main verb and object with idiomatic meaning(iii) N+V complex predicate.
Future Work
I Currently, we use argument structure clues for this distinction.Using this method [Arg par ’on’] zOr ’pressure’ dE ’give’might be a complex predicate because it has a par markedargument in place of canonical dative kO marked argument.
I However, the classification of [Arg kO ’on’] dhakkA ’push’ dE’give’ is more difficult as it has kO marked argument similar tothe canonical usage of dE ’give’.
References